LETTER doi:10.1038/nature10095 Hydrogen-poor superluminous stellar explosions R. M. Quimby1, S. R. Kulkarni1, M. M. Kasliwal1, A. Gal-Yam2, I. Arcavi2, M. Sullivan3, P. Nugent4, R. Thomas4, D. A. Howell5,6, E. Nakar7, L. Bildsten5,8, C. Theissen9, N. M. Law1,10, R. Dekany11, G. Rahmer11, D. Hale11, R. Smith11, E. O. Ofek11, J. Zolkower11, V. Velur11, R. Walters11, J. Henning11, K. Bui11, D. McKenna11, D. Poznanski4,12, S. B. Cenko12 & D. Levitan13 Supernovae are stellar explosions driven by gravitational or ther- monuclear energy that is observed as electromagnetic radiation emitted over weeks or more1. In all known supernovae, this radi- ation comes from internal energy deposited in the outflowing ejecta by one or more of the following processes: radioactive decay of freshly synthesized elements2 (typically 56Ni), the explosion shock in the envelope of a supergiant star3, and interaction between the debris and slowly moving, hydrogen-rich circumstellar material4. Here we report observations of a class of luminous supernovae whose properties cannot be explained by any of these processes. The class includes four new supernovae that we have discovered and two previously unexplained events5,6 (SN 2005ap and SCP 06F6) that we can now identify as members of the same class. These supernovae are all about ten times brighter than most type Ia supernova, do not show any trace of hydrogen, emit significant ultraviolet flux for extended periods of time and have late-time decay rates that are inconsistent with radioactivity. Our data require that the observed radiation be emitted by hydrogen-free material distributed over a large radius ( 1015 centimetres) and expanding at high speeds (.104 kilometres per second). These long-lived, ultraviolet-luminous events can be observed out to redshifts z . 4. The Palomar Transient Factory7,8 (PTF) is a project dedicated to finding explosive events and has so far identified over one thousand supernovae. PTF09atu, PTF09cnd and PTF09cwl (also known as SN 2009jh9) were detected using the Palomar Observatory’s 1.2-m Samuel Oschin Telescope during commissioning of the PTF system, in 2009, and PTF10cwr10–12 (SN 2010gx13) was detected the following year (Fig. 1; see Supplementary Information, section 1). As with other supernova candi- dates, optical spectra for classification were obtained using the W. M. Keck Observatory’s 10-m Keck I telescope, Palomar Observatory’s 5.1-m Hale Telescope, and the 4.2-m William Herschel Telescope. The spectra (Fig. 2) show broad absorption dips at short wavelengths and mostly smooth continua at longer wavelengths. We further identify narrow absorption features in the PTF spectra from the Mg II doublet at rest wavelengths 2,796 A˚ and 2,803 A˚ , and measure redshifts of z 5 0.501, 0.258, 0.349 and 0.230 for PTF09atu, PTF09cnd, PTF09cwl and Figure 1 | Ultraviolet-luminous transients discovered by the PTF. Left: before explosion; right: after explosion; top to bottom: PTF09atu, PTF09cnd, PTF09cwl and PTF10cwr. Each tile shows a false-colour image constructed by assigning image data from three separate band passes to red, blue and green (g, r and i bands, respectively, for PTF09atu; u, g or V, and r bands for PTF09cnd, PTF09cwl and PTF10cwr). In each case, Sloan Digital Sky Survey reference data form the pre-explosion image. The post-explosion images are composed from observations made with the Palomar Observatory’s 1.5-m telescope, the Wise Observatory’s 1.0-m telescope and the Ultraviolet/Optical Telescope on board NASA’s Swift satellite. 1Cahill Center for Astrophysics 249-17, California Institute of Technology, Pasadena, California 91125, USA. 2Benoziyo Center for Astrophysics, Faculty of Physics, Weizmann Institute of Science, 76100 Rehovot, Israel. 3Department of Physics (Astrophysics), University of Oxford, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK. 4Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, California 94720, USA. 5Las Cumbres Observatory Global Telescope Network, 6740 Cortona Drive, Suite 102, Goleta, California 93117, USA. 6Department of Physics, University of California, Santa Barbara, Broida Hall, Santa Barbara, California 93106, USA. 7Raymond and Beverly Sackler School of Physics & Astronomy, Tel Aviv University, Tel Aviv 69978, Israel. 8Kavli Institute for Theoretical Physics, Kohn Hall, University of California, Santa Barbara, California 93106, USA. 9University of California, San Diego, Department of Physics, 9500 Gilman Drive, La Jolla, California 92093, USA. 10Dunlap Institute for Astronomy and Astrophysics, University of Toronto, 50 St George Street, Toronto, Ontario M5S 3H4, Canada. 11Caltech Optical Observatories, California Institute of Technology, Pasadena, California 91125, USA. 12Astronomy Department, University of California, Berkeley, 601 Campbell Hall, Berkeley, California 94720, USA. 13Department of Physics, California Institute of Technology, Pasadena, California 91125, USA. 23 JUNE 2011 | VOL 474 | NATURE | 487 ©2011 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER PTF10cwr, respectively. After combining the three available spectra of section 5), implying that the internal energy must have been deposited the SCP 06F6 transient, we find that the data correlate to the PTF sample at a radius that is not much smaller than Rph. and may also show narrow Mg II absorption with redshift z 5 1.189 Integrating the rest frame g-band light curve and assuming no bolo- (Supplementary Information, section 4). Similarly to all PTF events, metric correction, we find that PTF09cnd radiated ,1.2 3 1051 erg. A SN 2005ap (z 5 0.283) shows a distinct W-shaped absorption feature similar analysis of the SCP 06F6 data gives a radiated energy of near rest wavelength 4,300 A˚. Although the broad spectral features of SN ,1.7 3 1051 erg. We also fit Planck functions to the ultraviolet and 2005ap are systematically shifted to higher speeds, the overall resemb- optical observations of PTF09cnd (Supplementary Fig. 1) and find an lance to the other PTF events is striking. The PTF discoveries bridge the approximate bolometric output of ,1.7 3 1051 erg. The derived black- redshift gap between SCP 06F6 and SN 2005ap and link these once body radii indicate a photospheric expansion speed of vph < 14,000 km disparate events, thus unifying them all into a single class. s21. If the main source of luminance were the conversion of kinetic With the redshifts above and a standard flat cosmology with Hubble energy, then the bolometric energy would require ,1M[ of material at parameter H0 5 71 and matter energy density Vm 5 0.27, the peak this speed, assuming a conversion efficiency of 100%. A more realistic absolute u-band AB magnitudes14 for the PTF transients in the rest efficiency factor would make the minimum mass a few times larger. frame are near 222 mag and that for SCP 06F6 is near 222.3 mag Because no traces of hydrogen are seen in any of the spectra (Fig. 3). The ,50-d rise of SCP 06F6 to maximum in the rest frame is (Supplementary Information, section 3), interaction with ordinary compatible with the PTF sample, although there seems to be some hydrogen-rich circumstellar material is ruled out. We thus conclude diversity in the rise and decline timescales. To power these high peak that these events cannot be powered by any of the commonly invoked 56 magnitudes with radioactivity, several solar masses (M[)of Ni are processes driving known supernova classes. needed (.10M[, following ref. 15), and yet in the restframe V band, the The early spectra presented here are dominated by oxygen lines and post-maximum decline rates of the PTF events are all .0.03 mag d21, do not show calcium lines, iron lines or other features commonly seen which is a few times higher than the decay rate of 56Co (the long-lived in ordinary core-collapse supernovae. The lack of metals is particularly daughter nucleus of 56Ni). These are therefore not radioactively powered noticeable in the ultraviolet flux, which is typically depleted by absorp- events. tion. These events are hosted by low-luminosity galaxies that may pro- Next we check whether the observed photons could have been vide a subsolar progenitor environment (Supplementary Information, deposited by the explosion shock as it traversed the progenitor star. section 6). The new class of events we have identified is thus obser- The photospheric radius, Rph, that we infer for PTF09cnd at peak vationally characterized by extreme peak luminosities, short decay luminosity, on the basis of the observed temperature and assuming times inconsistent with radioactivity, and very hot early spectra with 15 blackbody emission, is Rph < 5 3 10 cm (Supplementary Fig. 1). If significant ultraviolet flux and lacking absorption lines from heavy the radiated photons were generated during the star explosion, then elements such calcium and iron, which are commonly seen in all other adiabatic losses would result in only a fraction R*/Rph of the energy types of supernova. remaining in the radiation at any given time, where R* is the initial These observations require a late deposition of a large amount of stellar radius. Given that the energy radiated around the peak is energy (.1051 erg) into hydrogen-poor, rapidly expanding material 51 23 ,10 erg and that R*/Rph , 10 for virtually any hydrogen-stripped (slow-moving material would produce narrow spectroscopic features, progenitor, this model requires an unrealistic total explosion energy of which are not observed). We point out two possible physical processes .1054 erg. In fact, the large radius and the duration of PTF09cnd leave that can perhaps power these superluminous sources. One is a strong almost no place for adiabatic losses (Supplementary Information, interaction with a massive, rapidly expanding, hydrogen-free shell. 10 b O II Si III Si III Mg II C II 1 ) a c −1 Å –2 SN 2005ap cm 10.0 z = 0.283 −1 PTF10cwr z = 0.230 PTF09cnd 1.0 z = 0.258 Scaled flux (erg s Scaled flux (erg PTF09cwl z = 0.349 PTF09atu 0.1 z = 0.501 SCP 06F6 Mg II z = 1.189 2,000 4,000 8,000 2,800 2,900 Rest wavelength (Å) Figure 2 | Spectral energy distributions of the SN 2005ap-like sample.